We're getting 502 during pod deployment. From what I can see, this is due to Traefik forwarding requests to a pod that's received it's shutdown signal and therefore no longer accepts traffik.
We're using regular Ingress objects (not CRDs), running the latest Traefik version.
Have anyone else seen this behavior?
Hello @trond
Thank you for your interests in Traefik.
Traefik relies on Kubernetes health probes to decide whether backend is healthy and ready to accept incoming forward network traffic to the pod.
If the probes defined on a pod level are throwing an error issues the Kubernetes endpoint will not be created. It means that Traefik will not forward the request because there will not be the endpoint.
I would check if the probes are added to your deployment, what is the deployment strategy and its configuration.